Please can someone help?--negative AIC, BIC, CAIC and AIC-3 values when runing FIMIX-PLS

[xiaoyu chen]

Hi,
I'm trying to detect the un-observed heterogeneity behind my research model. When runing the finite mixture (FIMIX) segmentation using Smartpls software 3.0, I got negative AIC, BIC, CAIC values (in fact, most of the quality criteria listed by Smartpls were negative). Meanwhile, I found that:

1) The absolute values of AIC, BIC, and CAIC were smaller when number of segment K=2 than when K=3. (e.g. AIC = -9048.191 when K=2,  AIC=-9210.889 when K=3  ) .

2) The EN value was 1 when the number of segment K=2,  while EN value reduces to 0.763 when K=3.

When I read journal papers using FIMIX segmentation method, it seems they all report positive AIC, BIC, CAIC and AIC-3 values? Does that means I need use absolute values of AIC, BIC and CAIC ?  Is EN=1 credible?

It will be highly appreciated if someone can help answer these questions in my mind.

Kind regards

Xiaoyu Chen